PEM's New Pharmaceutical Case Management Model For Medicaid Patients

Transcription

1 Evaluation of the Iowa Medicaid Pharmaceutical Case Management Program Elizabeth A. Chrischilles, Barry L. Carter, Brian C. Lund, Linda M. Rubenstein, Shari S. Chen-Hardee, Margaret D. Voelker, Tae-Ryong Park, and Angela K. Kuehl ABSTRACT Objective: To test the effect of pharmaceutical case management (PCM) on medication safety and health care utilization. Design: Prospective cohort design with 9-month follow-up period (enrollment from October 1, 2000, through July 1, 2001, with follow-up through July 1, 2002). Setting: Iowa Medicaid program. Participants: 2,211 noninstitutionalized, continuously eligible Iowa Medicaid patients taking four or more chronic medications including at least one agent commonly used in at least 1 of 12 specific diseases who were cared for by pharmacists in 117 pharmacies. Interventions: Reimbursement for PCM services (initial patient assessment, written recommendations to physician, follow-up assessments and communication of progress and new problems to physician). Main Outcome Measures: Use of high-risk medications, Medication Appropriateness Index (MAI) score, health care utilization. Results: Pharmacists in 114 pharmacies had eligible patients during at least one quarter during the study period; 28 pharmacies were classified as high intensity based on the number of PCM patients they managed. A total of 524 of the eligible patients received 1,599 PCM services; 90% of claims were filed by pharmacists, and the remainder by physicians. Nearly one half (46.1%) of medications and 92.1% of patients had at least one medication problem before PCM. By closeout, the percentage of medications with problems decreased in 8 of 10 MAI domains for those who received PCM. Compared with baseline, mean MAI score improved significantly from 9.4 to 8.3 among PCM recipients (P <.001). Percentage of PCM recipients using high-risk medications decreased significantly compared with PCM eligibles who did not receive the service. In the 28 pharmacies that adopted the new service most intensely, patients had a significant decrease in high-risk medication use, compared with patients of low-intensity pharmacies (P <.001). No difference was observed between PCM recipients and PCM eligibles who did not receive PCM in health care utilization or charges, even after including reimbursements for PCM. Conclusion: Medication safety problems were prevalent in this high-risk population. The PCM program improved medication safety during a 9-month follow-up period. Keywords: Medicaid, polypharmacy, pharmaceutical care, pharmacotherapy, health resource utilization, economics. J Am Pharm Assoc. 2004;44: Received June 9, 2003, and in revised form September 24, Accepted for publication November 15, Elizabeth A. Chrischilles, PhD, is Director, Health Effectiveness Research Center, and Professor, Department of Epidemiology, College of Public Health, University of Iowa, Iowa City. Barry L. Carter, PharmD, FCCP, FAHA, is Professor and Head, Division of Clinical and Administrative Sciences, and Professor, Department of Family Medicine, College of Medicine, University of Iowa, Iowa City. Brian C. Lund, PharmD, MS, is Associate Director of Research, Laureate Psychiatric Research Center, Tulsa, Okla.; at the time of this study, Dr. Lund was Postdoctoral Research Scholar, Colleges of Public Health and Pharmacy, University of Iowa, Iowa City. Linda M. Rubenstein, PhD, is Senior Statistician; Shari S. Chen-Hardee, MS, is Senior Data Analyst; Margaret D. Voelker, PhD, is Epidemiologist and Assistant Director; Tae-Ryong Park is Graduate Research Assistant; and Angela K. Kuehl, PharmD, MS, is Clinical Pharmacist and Senior Research Assistant, Health Effectiveness Research Center, Department of Epidemiology, College of Public Health, University of Iowa, Iowa City. Correspondence: Elizabeth A. Chrischilles, PhD, Department of Epidemiology, College of Public Health, University of Iowa, Iowa City, IA Fax: e-chrischilles@uiowa.edu Disclosure: The authors declare no conflicts of interest or financial interests in any product or service mentioned in this article, including grants, employment, gifts, stock holdings, or honoraria. Funding Sources: For Pharmaceutical Case Management Services, State of Iowa General Assembly and federal matching funds for Iowa Medicaid; for the evaluation of the service reported here, State of Iowa General Assembly, the Iowa Pharmacy Foundation, McKesson HBOC, and the Institute for the Advancement of Community Pharmacy. Acknowledgments: To Thomas Temple and Nancy Bell, Iowa Pharmacy Association; Cheryl Clarke, David Scholz, Sue Carter, Randal McDonough, William Doucette, Karen Farris, Jay Currie, Michael Ernst, William Miller, Gary Rosenthal, Joseph T. Hanlon, and all the Iowa pharmacists and physicians who participated in this program. See related articles on pages 350 and 358 of this issue. Presented previously at the Iowa Pharmacy Expo, Des Moines, Iowa. January 19, Vol. 44, No. 3 May/June Journal of the American Pharmacists Association 337

2 Iowa Pharmaceutical Case Management Medication safety has become a major concern in the United States. 1 The rate of adverse drug events increases substantially as the number of drugs in a regimen increases. 2,3 Published evidence strongly suggests that pharmacist physician teams can increase medication safety Recently, Gurwitz et al. 19 concluded from their prospective cohort study of more than 30,000 Medicare enrollees that adverse drug events are common and preventable in the community. They specifically recommend that further development and testing of new approaches to enhance collaborations between those who prescribe drugs and those who know the most about the specific drugs, that is, clinical pharmacists, should be pursued in the ambulatory setting. Numerous examples of innovative practice models in community pharmacy have emerged during the past 35 years. 4 Studies in community pharmacies have demonstrated that interventions and management by pharmacists can improve control of several chronic conditions, provide preventive services, and provide interventions that can significantly reduce the cost of therapy Several demonstration projects and research studies in community pharmacy have been conducted in Iowa. 12,13,20,21 The Iowa AT A GLANCE Synopsis: The state-mandated evaluation of the initial experiences of pharmacists, physicians, and patients in the Iowa Medicaid Pharmaceutical Case Management (PCM) program is presented in this article. To improve medication effectiveness and safety, the PCM program reimburses pharmacists and physicians for interventions, monitoring, and follow-up in eligible Medicaid recipients. Medication appropriateness increased significantly among the PCM recipients, compared with eligible recipients who did not receive these interventions, and use of high-risk medications decreased significantly. Patients of pharmacies that adopted the program most intensely had a significant decrease in high-risk medication use. No differences in health care utilization or charges were found between PCM-eligible patients who received services and those who did not, even with charges for PCM services included. Analysis: Medication-related problems are of great concern in the United States, and this study documents that collaborative efforts that link motivated pharmacists, caring physicians, and a payment mechanism can improve drug therapy in at-risk patients. Developing and sustaining PCM services in community pharmacies is challenging, but the beneficial effect observed among patients receiving these services provides strong support for this model in other states and for at-risk patients whose care is covered by other payers. Center for Pharmaceutical Care program has trained more than 200 practicing pharmacists in strategies to identify and resolve drug-related problems. 12,13,20,21 The Iowa Medicaid Pharmaceutical Case Management (PCM) program is the first such program in the United States that formally recognizes pharmacists as providers and pays physician pharmacist teams who deliver medication management services for high-risk patients in the community setting. The Iowa Medicaid PCM program began with funds appropriated during the 2000 Iowa legislative session. The program was implemented as a state plan amendment and is an ongoing service for eligible patients receiving Medicaid. The program was designed to benefit a subset of individuals who are at very high risk for having adverse effects from their medications. The design, fee structure, pharmacist qualifications, advisory board, and services provided during the first 3 months of patient enrollment have been previously published 22 and are reviewed only briefly here. The unique features of the PCM program are that both physicians and pharmacists are paid for their services and that a collaboratively determined action plan can be implemented by the pharmacist without requiring a patient visit to a physician. Under this initiative, both pharmacists and physicians provide and may be paid for one initial assessment ($75 each), up to four problem follow-up assessments ($40 each) per 12 months; up to two new problem assessments ($40 each) per 12 months; and up to one preventive follow-up assessment ($25 each) every 6 months. Eligible patients (noninstitutionalized Medicaid beneficiaries of any age who are taking at least 4 medications and with one of 12 specified diseases) receive an initial assessment by the pharmacist who makes recommendations to the physician in writing. Recommendations that are accepted or modified by the physician are considered an action plan. Pharmacists make problem followup assessments until all problems are resolved, communicating with the physician in each case. Once problems are resolved, preventive follow-up assessments can occur every 6 months, and new problems that arise episodically can trigger a new problem assessment and a new action plan. The aim of the Iowa Medicaid PCM program is to avoid adverse drug events and the health system costs associated with these adverse effects. The means to accomplish this is by using more optimal, lower-risk medication regimens. The legislation that implemented the Iowa Medicaid PCM program also required program evaluation, and the results of that research are reported in this article. Objectives The primary objectives of this evaluation were to describe the extent and content of PCM services and determine the effect of the PCM program on medication safety. A secondary objective was to estimate the effect of PCM services on health care utilization. 338 Journal of the American Pharmacists Association May/June 2004 Vol. 44, No. 3

3 Iowa Pharmaceutical Case Management RESEARCH Methods Study Population The design for the evaluation of the PCM program has been previously described. 22 A prospective cohort design was used and included patients who became eligible for PCM from October 1, 2000, through July 1, 2001, with follow-up through July 1, Thus the evaluation reports mainly on the PCM program start-up phase. The program was implemented with 117 participating pharmacies on October 1, Eligible patients were identified quarterly using pharmacy claims data from participating pharmacies. Noninstitutionalized Iowa Medicaid patients taking four or more long-term medications, including at least one medication representing 1 of specified 12 diseases (n = 3,037), were eligible (the 12 diseases were congestive heart disease, ischemic heart disease, diabetes mellitus, hypertension, hyperlipidemia, asthma, depression, atrial fibrillation, osteoarthritis, gastroesophageal reflux, peptic ulcer disease, and chronic obstructive pulmonary disease).to have complete claims data, we excluded from analysis all patients who were not continuously eligible for Medicaid from 6 months before through 12 months after the date on which they became eligible for PCM. Data Sources and Variable Definitions Data sources included Medicaid claims files, mailed questionnaires to eligible patients at baseline (mailed before the first day of eligibility for receipt of PCM services), quarterly fax surveys of all participating pharmacies, and review of problem-oriented patient records kept in pharmacies for recipients of the service. Variables constructed from these sources are described in the following sections. Measures from Medicaid Claims Files Receipt of PCM services was measured as at least one claim for PCM reimbursement filed within 9 months of the patient s PCM eligibility. Medication safety measures were constructed based on an active drug list constructed from pharmacy claims. A computer algorithm was developed to construct, from Medicaid pharmacy claims, a list of drugs considered active on the date a patient became eligible for PCM (the index date). 23 This same process was conducted 9 months later (i.e., on the follow-up date). Two measures of medication safety were applied to these drug lists. The first measure was a rating of medication problems for all patients who received PCM services, which was constructed from a clinical pharmacist conducted structured review of patients pharmacy records (see Measures From Patients Pharmacy Records below). The second measure was use of medications considered to be inappropriate for use among persons aged 65 and older (hereafter referred to as high-risk medication use). 24 This measure was constructed for all PCM-eligible elders (aged 65 and over), inclusive of those who did and did not receive the services. Measures of health care utilization included number of claims and charges to Medicaid for pharmacy, institutional, and medical services. The PCM claims from both pharmacies and physicians were included in the medical claims analysis (because they were submitted on a Health Care Financing Administration [HCFA- 1500] claim form, they reside in the medical claims file). Measures from Mailed Baseline Patient Questionnaires Prevalence of adverse drug reactions was assessed in the patient questionnaires using the question, In the past year, have you noticed any side effects, unwanted reactions, or other problems from medications you were taking? 2 Generic health status was measured on the questionnaire with the Medical Outcomes Study Short Form (SF-36), using the physical and mental health summary scores. 25 Measures from Quarterly Pharmacy Fax Survey Three months after receiving their list of eligible patients, pharmacies received fax surveys querying the status of eligible patients, and these were returned for 96.7% of eligible patients. These data were used to measure the intensity with which a pharmacy provided PCM. 22 The percentage of PCM-eligible patients for whom a pharmacist recommendation was completed was calculated each quarter for each pharmacy as a measure of the intensity of PCM adoption. A pharmacy that completed recommendations for at least 50% of patients in at least one quarter was considered a high-intensity pharmacy. Moderate-intensity pharmacies were those that completed recommendations for at most 25% to 49% of patients in any quarter, low-intensity pharmacies completed recommendations for 1% to 24% of patients, and zero-intensity pharmacies completed no recommendations during the study period. Measures from Patients Pharmacy Records Pharmacists who provide PCM are required to maintain documentation of all PCM services in problem-oriented notes in the SOAP format (subjective, objective, assessment, plan). Copies of these records were obtained for those who received PCM services 1 year after each patient s initial PCM eligibility date. These records served as the source of detailed information about medical diagnoses, medication purpose, and dosage, which were required for construction of a complete Medication Appropriateness Index (MAI) score for patients who received PCM services. MAI scores are derived by rating each medication using 10 weighted explicit criteria that are classified by the reviewer as either appropriate, marginally appropriate, or inappropriate, on the basis of strict operational definitions for each criterion. 26,27 The 10 criteria that contribute to the MAI score are: indication, effectiveness, correct dosage, correct directions, practical directions, drug drug interaction, drug disease interaction, duplication, duration of treatment, and cost. The MAI score for a medication can range from 0 to 18 (higher is more inappropriate). Patient-specific scores depend on the number of medications rated so both the summed MAI score Vol. 44, No. 3 May/June Journal of the American Pharmacists Association 339

4 Iowa Pharmaceutical Case Management and the mean MAI score (i.e., the average MAI rating for all medications prescribed) were examined. 27 The MAI was scored using the weights supplied by the developer; items coded as inappropriate ( C rating) were assigned weights. 26 The MAI was calculated from problem-oriented patient records kept by pharmacists, which has previously been demonstrated to be reliable in community pharmacy settings. 28 MAI scores were determined by a clinical pharmacist blinded to PCM intensity and pharmacy. The change in full MAI score from the index date (before PCM) to the 9-month follow-up date was evaluated for all patients who received the service. In addition to the MAI reviews, a 40% random sample of these records was abstracted to describe the action plans developed by the care teams. Statistical Methods To determine the effects of the PCM program, three types of comparisons were made: 1. Among patients who received PCM services, medication safety on the index date was compared with safety of medications on the 9-month follow-up date (i.e., using a single group pre post design). 2. Between PCM-eligible patients who received PCM services and PCM-eligible patients who did not receive PCM services, the change over time in use of high-risk medications, number of active drugs, cost of their medications, and medical and institutional utilization (i.e., a comparison group design). 3. Among PCM-eligible patients, regardless of receipt of services, the effect of pharmacy intensity on change in use of high-risk medications and on change in health care utilization. The relationships of continuous variables at baseline were assessed using t-tests, analysis of variance (ANOVA), analysis of covariance (ANCOVA) (controlling for age and gender), and nonparametric procedures. Correlation analyses used Pearson or Spearman methods. Comparisons between categorical variables were assessed using chi-square statistics or exact nonparametric methods for small sample sizes. Cochran Mantel Haenszel methods were used when controlling for age, gender, and other characteristics. The relationship of continuous and categorical variables was assessed with Wilcoxon/Kruskall Wallis methods, ANOVA, and ANCOVA (controlling for age and gender). Longitudinal statistical analysis of change over time in medication safety and health care utilization used the generalized estimating equations (GEE) approach. 29 This method accounts for correlations between repeated measures. GEE supports generalized linear models for a variety of longitudinal or correlated data and can be used for Poisson, logistic, gamma, and normal distribution analyses. The logistic function was used for binomial outcomes, the cumulative logistic function for multinomial outcomes, and the normal distribution for continuous outcomes. Binomial variables included individual categories of medication safety problems such as the 10 domains for the MAI score and use/nonuse of high-risk medications. Multinomial outcomes included number of emergency department claims per month and number of inpatient claims per month. Continuous outcomes included number of drugs, total MAI scores, number of outpatient claims per month, number of medical service claims per month, and log charges. Models with dollar amounts used log dollars. Repeated measures analyses for repeated observations over time used the patient identification number as the unit of repetition. Results Description of Eligible Patients A total of 3,037 patients were eligible for PCM services during the study enrollment year (Table 1). The mean (± SD) age was 52.5 ± 20.2 years. Almost two thirds of eligible patients were aged 45 or older, 28.3% were aged 65 and older, and 6.4% were children. Adults ranged from 18 years to 101 years of age. Overall, 71.4% of patients were women. For patients who became eligible for PCM services during the four calendar quarters of the study beginning October 1, 2000, January 1, 2001, April 1, 2001, and July 1, 2001 PCM claims had been submitted by May 31, 2002 for 690 patients (22.7% of 3,037 eligible patients; Table 1) and 1,599 services. Of the 1,599 PCM services reimbursed, 90% (n = 1,440) were submitted on claims from pharmacists and 159 were from physicians. Among the 3,037 patients who were eligible for PCM, 2,211 (72.8%) remained continuously eligible for Medicaid from 6 months before through 12 months after the date at which they became eligible for PCM services and constituted the analysis dataset (Table 1). Of these, 524 patients received PCM services. A total of 659 patients including 180 who received PCM and 479 who did not responded to the baseline questionnaire (Table 1). Of 117 eligible pharmacies statewide, 114 had eligible patients assigned in at least one quarter. Of these, 28 pharmacies were classified as high intensity, 46 as moderate intensity, 21 as low intensity, and 19 as zero intensity. Table 2 displays the baseline (before Table 1. Data Availability for PCM-Eligible Patients Characteristic No. Patients Eligible to receive PCM 3,037 Received PCM 690 Continuously eligible for Medicaid a 2,211 Received PCM 524 Responded to questionnaire 180 Pharmacy record available 507 Did not receive PCM 1,687 Responded to questionnaire 479 Pharmacy record available 0 Abbreviation used: PCM, pharmaceutical case management. a Remained continuously eligible for Medicaid from 6 months before through 12 months after the date at which they became eligible for PCM services. 340 Journal of the American Pharmacists Association May/June 2004 Vol. 44, No. 3

5 Iowa Pharmaceutical Case Management RESEARCH Table 2. Baseline Characteristics of PCM-Eligible Patients, Adjusted for Age and Gender Patient Received PCM Patient Did Not Receive PCM (n = 524) (n = 1,687) Characteristic Mean ± SE Mean ± SE Mean age in years (± SE) a 54.1 ± ± 0.5 Mean (± SE) number of drug products a 7.5 ± ± 0.1 Mean (± SE) number of ingredients 8.3 ± ± 0.1 Mean (± SE) SF-36 physical summary score 33.1 ± 0.9 (n = 180) 32.8 ± 0.5 (n = 479) Mean (± SE) SF-36 mental summary score 42.5 ± 0.8 (n = 180) 42.2 ± 0.5 (n = 479) No. Patients (%) No. Patients (%) Gender no. women (%) a 419 (80.0) 1,169 (69.3) Ethnic background White 467 (89.1) 1,519 (90.0) Black 31 (5.9) 93 (5.5) Other 5 (1.0) 35 (2.1) Unknown 21 (4.0) 40 (2.4) Health status Hospitalizations in last year 47 (9.0) 202 (12.0) Emergency department visits in last year 105 (20.0) 423 (25.1) Baseline medication problems Drug drug interactions a,b 30 (5.7) 52 (3.1) Therapeutic duplications c 210 (40.1) 686 (40.7) Contraindicated drugs d in patients age 65 and older 76 (43.4) 131 (35.8) High dosage e error 88 (16.8) 231 (13.7) Any problem 295 (56.3) 875 (51.9) Adverse drug reactions in past 12 months 45 (25.0% of 127 (26.5% of 180 survey respondents) 479 survey respondents) Abbreviations used: PCM, pharmaceutical case management; SE = standard error; SF-36, Medical Outcomes Study Short Form. a P-value.05 for difference between received PCM and no PCM, all means adjusted for age and gender, using generalized linear models for continuous variables and logistic models for binomial variables. The dependent variable was the variable of interest at baseline. b Based on comparison of index date active drug lists with a table of level 1 or 2 drug drug interactions as listed in reference 41. c Therapeutic duplication was a finding of two or more drugs in the index date active drug list in the same therapeutic category as defined by Hanlon et al. in reference 3. d As defined by Beers in reference 24. e Dose of drug greater than that approved based on adult or geriatric maximum dose values, per patient age. PCM) sociodemographic, health status, and medication characteristics of those who received PCM services compared with those who were eligible for but did not receive PCM services, adjusted for between-group differences in patient age and gender. After adjusting for age differences, those who received PCM still took a higher number of medications and were more likely to be women and to have drug drug interactions. The types of drugs taken by those who did and did not receive PCM services were similar. Regardless of whether they received PCM services, more than one half of PCM-eligible patients had at least one indicator of inappropriate medication use during the baseline (pre-pcm) period. After adjusting for age and gender, at baseline those who received PCM services were similar to those who did not in overall, physical, and mental health, prior use of urgent care services, and prevalence of adverse drug reactions. Their health status, as measured using the 36-item short-form health survey (SF-36), was markedly lower than the national average for physical health (33 for PCM versus 50 for the United States, or 1.7 SDs below the national mean) and mental health (43 for PCM versus 50 for the United States, or 0.7 SDs below the national mean). Approximately 11.3% (249 of 2,211 patients) had been hospitalized and 23.9% (528 of 2,211 patients) had used an emergency department in the year before eligibility for PCM services. Of 659 survey respondents, 172 (26%) reported an adverse drug reaction Vol. 44, No. 3 May/June Journal of the American Pharmacists Association 341

6 Iowa Pharmaceutical Case Management Table 3. Recommendations for a Random Sample of Patients Who Received PCM Services No. No. Patients Recommendations Recommendation Type (%; n = 203) a (%; n = 771) Discontinue medication 67 (33.0) 106 (13.7) Start medication 105 (51.7) 189 (24.5) Change medication 73 (36.0) 105 (13.6) Change dose 83 (40.9) 125 (16.2) Change route 0 (0) 0 (0) Change schedule 28 (13.8) 35 (4.5) Change dosage strength 4 (2.0) 4 (0.5) Change dosage form 5 (2.5) 5 (0.7) Change treatment duration 0 (0) 0 (0) Therapeutic/disease state monitoring 78 (38.4) 117 (15.2) Enhance compliance 2 (1.0) 2 (0.3) Patient education 40 (19.7) 59 (7.7) Provider education 14 (6.9) 19 (2.5) Unspecified 3 (1.5) 5 (0.6) a Several recommendations were made per patient, and numbers therefore do not sum to 203. in the year before PCM services, a rate 2.6 times that observed using the same question in a survey of a population-based sample of elderly Iowans. 2 Categorizing Pharmacists Recommendations A random sample of 203 patient pharmacy charts was reviewed to characterize the nature of the problems identified, recommendations made by pharmacists, and physician acceptance of these recommendations. The communication form between the pharmacist and the physician was the source used to identify recommendations. Pharmacists identified a mean (± SD) of 2.6 ± 1.6 (median 2.0, range 1 9) medication problems per patient and made 3.8 ± 3.0 (median 3.0, range 1 15) recommendations per patient. The most common types of problem detected were: therapeutic monitoring needed (41.9% of patients, 16.7% of recommendations); untreated conditions (40.4% of patients, 17.6% of recommendations); and undertreated conditions (37.0% of patients, 14.3% of recommendations). The most common type of recommendation made was to start a new medication (Table 3; 51.7% of patients, 24.5% of all recommendations). Other common recommendations were to change the dose of a medication, change a medication to an alternate therapy, monitor the medicine or disease (e.g., monitor drug levels or blood pressure), or to discontinue a medication. Physicians accepted a mean (± SD) of 1.9 ± 2.0 (median 2.0, range 0 15) recommendations per patient. Of 771 recommendations made by pharmacists, 379 (49.2%) were accepted by physicians. Physicians required a mean of 8.9 days (median, 4.0 days) to relay their decision about pharmacists recommendations. Effect of PCM on Medication Safety Change Over Time Among Those Who Received PCM Services A pharmacy record was available for 507 of 524 patients who had received PCM (Table 1). Pharmacy records, and hence MAI scores, were not available for those who did not receive the service. Changes in MAI scores, number of medications, and individual medication problems between the index date (the date a patient became eligible for PCM services) and the follow-up date (9 months later) were analyzed using GEE for repeated measures. We used normal regression methods when the dependent variable was a total score and logistic methods when the dependent variable was binary (yes/no). Independent variables were age (continuous), gender, pharmacy intensity, a time indicator (1 if index date; 2 if follow-up date), and a time by pharmacy intensity interaction term. Table 4 displays the medication problems and MAI scores at the index date and the follow-up date. Overall, nearly one half (46.1%) of medications and 92.1% of patients had at least one medication problem on the index date (pre-pcm). By closeout, the percentage of medications with problems decreased in 8 of 10 MAI domains for those who received PCM. Overall, the percentage of medications with any problems decreased from 46.1% to 40.9% (P <.001). No significant time by intensity interactions were identified. This indicates that the change in MAI measures for patients who received PCM did not vary by PCM intensity of the pharmacy. The mean number of ingredients increased significantly from 7.9 to 8.2 among those who received PCM and the mean MAI score decreased significantly from 9.4 to 8.3, a 12% improvement (P <.001). After receiving PCM services, patients were significantly less likely to be taking a drug without a valid indication for use or that was considered ineffective or inappropriate for age (effectiveness), had an incorrect dosage, interacted with a patient disease, or had an inappropriate duration of use. Though not statistically significant, a trend was observed for patients to be less likely to be taking a drug that had incorrect directions or unjustifiable expense. These results indicate that medication use improved significantly from before to 9 months after patients received PCM services. This was despite an increase in mean number of active ingredients from baseline to follow-up (Table 4). Though the MAI evaluation provided a detailed assessment of change in the domains of medication appropriateness, the single group pre post comparison is not a strong design for evaluating the impact of the intervention. A comparison group design was possible for one domain of medication safety and also for analysis of health care utilization effects. Change in Use of High-Risk Medications To compare change in medication problems for patients who received PCM services with those who were eligible but did not receive them, we used the list of medications established through a consensus-forming process by Beers 24 to be inappropriate for people aged 65 and over. Other pharmacy claims-based measures 342 Journal of the American Pharmacists Association May/June 2004 Vol. 44, No. 3

7 Iowa Pharmaceutical Case Management RESEARCH Table 4. MAI Ratings from Index Date and at Follow-Up for Patients Who Received PCM Services a Medications (n = 8,142) Patients Index Date Follow-up Index Date Follow-up (n = 4,001) (n = 4,141) (n = 507) (n = 505) No. (%) No. (%) No. (%) No. (%) With With With With Medication Problem (Weight) Total b Problem Total b Problem P-value c Total b Problem Total b Problem P-value c Indication (3) 3, (3.3) 3, (2.5) < (18.6) (14.4).002 Effectiveness (3) 3, (6.8) 3, (5.1) < (35.0) (29.9).001 Correct dosage (2) 3, (9.8) 3, (7.3) < (46.8) (40.5) <.001 Correct directions (2) 3, (10.2) 3, (9.0) (48.4) (44.4).056 Practical directions (1) 3, (6.9) 3, (5.9) (33.7) (32.5).785 Drug drug interaction (2) 3, (0.9) 3, (0.9) (5.8) (6.8).295 Drug disease interaction (2) 3, (6.8) 3, (5.8) (36.5) (32.9).014 Duplication (1) 3, (11.3) 3, (9.8) (53.1) (51.3).400 Duration of treatment (1) 3, (9.6) 3, (8.0) (43.3) (39.1).006 Cost (1) 3, (16.7) 3, (15.9) (67.1) (65.2).316 At least one of the above problems 3,646 1,681 (46.1) 3,791 1,552 (40.9) < (92.1) (89.1).008 MAI Descriptive Statistics Total Statistic Total Statistic Total Statistic Total Statistic Mean (± SD) ingredients d ± ± Median ingredients d Range in ingredients d Mean (± SD) MAI score e 3, ± 1.8 3, ± 1.6 < ± ± 7.1 <.001 Median MAI score 3, , Range in MAI score 3, , Abbreviations used: MAI, Medication Appropriateness Index; SD, standard deviation; PCM, pharmaceutical case management. a The index date was the day first eligible for PCM: follow-up was 9 months after the index date. b Because of incomplete information in some records, the number of medications rated was less than the total number of medications (n = 4,001 at index date and n = 4,141 at follow-up) and the number of patient charts rated was less than the total patient charts available (n = 507 at index date and n = 505 at follow-up). c P value was for difference between index and follow-up dates based on generalized estimating equations for repeated measures with two time points for continuous ( normal ) and binomial variables. d Ingredients were the active components of drug products; some combination products contained multiple active ingredients. e The MAI score was calculated by summing the weight for each MAI problem for each active ingredient. The mean MAI score per medication was the total of the summated MAI scores divided by the total number of active ingredients rated. The mean MAI score per patient was the total of the summated MAI scores (excluding patients with missing data for any MAI problem) divided by the number of patients (n = 471 index date, n = 469 follow-up). were tested and found to have poor validity (kappa coefficients of 0.28 or lower for agreement with clinician record-based review). Clinician review of patient records to identify use of inappropriate drugs (Beers list) agreed well with the computerized review of pharmacy claims data (kappa = 0.76). To test whether those who received PCM differed in rate of change from those who did not receive PCM, we used a GEE repeated-measures logistic model with medication safety (yes/no) as the dependent variable, controlled for age and gender, and included a time by PCM interaction in the model. The monthly repeated measure of use of high-risk drugs (yes/no) was calculated using the active drug list for the first of each month, from 3 months before through 9 months after the index date. Only patients aged 65 and older were included in the high-risk drug use analyses. A significant time by intervention interaction (P =.032) indicated that, after adjusting for age and gender, PCM recipients had a significantly greater decrease in use of high-risk medications than did PCM-eligible patients who did not receive the service. Figure 1 shows that the decline in high-risk medication use (Beers list) occurred after patients became eligible for PCM. We also compared the change from index date to follow-up date in prevalence of high-risk drugs within the two study groups. This model used only one prevalue and one postvalue (index date and follow-up date, respectively) in a GEE repeated-measures logistic methods analysis, controlled for age and gender, and tested for a Vol. 44, No. 3 May/June Journal of the American Pharmacists Association 343

8 Iowa Pharmaceutical Case Management Figure 1. Proportions of patients aged 65 and older taking a Beers high-risk medication, from 3 months before PCM eligibility through 9 months after becoming eligible, by whether PCM services were received (n = 175) or not (n = 366) Percent The time by interaction P value is Index Date Study Month (Pre-PCM months 1 3; PCM-eligible 5 13) PCM No PCM time by PCM difference. Those who received PCM had a significant (P <.05) decrease in prevalence of high-risk medication use (from 43.4% to 32.6%). In contrast, no significant change occurred among those who were eligible for but did not receive PCM services (from 35.8% to 34.4%). Interestingly, patients who received PCM services had a higher baseline prevalence of using these drugs than did patients who did not receive PCM services, suggesting that pharmacists may have targeted patients at higher risk. To evaluate whether there was a dose response relationship between the intensity with which the service was adopted by the pharmacy and the magnitude of improvement in medication safety, we examined the effect of pharmacy intensity on change in use of high-risk medications over time for all PCM-eligible patients. Because low-intensity pharmacies provided PCM to so few patients, the effect of PCM should not have been detectable in the patient population of these pharmacies (i.e., the intervention effect would have been diluted by the large number of patients who did not receive the intervention). In contrast, for high-intensity pharmacies where the majority of PCM-eligible patients actually received the service, the effect of PCM was expected to be detectable. We therefore hypothesized that there would be a significant time by pharmacy intensity interaction, specifically that medication safety would improve in high-intensity pharmacies to a greater extent than it would in low-intensity pharmacies. To test this hypothesis we used a GEE repeated-measures logistic model with high-risk medication use (yes/no) as the dependent variable at two time points (index date and follow-up date), controlled for age and gender, and looked for a significant time by PCM intensity interaction. We also tested for a significant time effect within each of the four intensity groups. As displayed in Figure 2, this hypothesis was supported. A significant P value (P =.005) was found for the time by intensity interaction, indicating that patients of higher-intensity pharmacies had a greater decrease in high-risk medication use. In addition, the decrease over time among high-intensity pharmacies was significant (P <.001), whereas for other intensity groupings it was not. PCM-eligible patients receiving prescriptions at high-intensity pharmacies had a higher baseline prevalence of medication problems than did patients receiving prescriptions from lower-intensity pharmacies (P <.05). Effect of PCM Services on Medication Use Adjusted for age and gender differences, when compared with PCM-eligible patients who did not receive PCM, PCM services 344 Journal of the American Pharmacists Association May/June 2004 Vol. 44, No. 3

9 Iowa Pharmaceutical Case Management RESEARCH Figure 2. Changes in percentage of PCM-eligible patients aged 65 and older taking high-risk medications from date of initial PCM eligibility (index date) to 9 months later (follow-up date), by pharmacy intensity. The differences shown were significantly different in high-intensity pharmacies and not significantly different in other categories of pharmacies Percent High Moderate Low Zero Intensity Index Date Follow-up Date had no significant effect on the net number of medications or medication charges (Table 5). For these models we used GEE repeated-measures normal regression methods with number of medications or charges as the dependent variable at two time points (index date and follow-up date). We controlled for age and gender and tested for a significant time by PCM interaction to evaluate change over time. The number of drugs and charges tended to increase both for those who did and did not receive PCM services. Effects of PCM on Resource Utilization We used several different models to examine the effect of PCM services on inpatient, outpatient, medical service, and emergency department utilization and charges. All models included 18 months of information for utilization (number of claims) or charges (dollars). Independent/control variables included age (continuous), gender, pharmacy intensity (high versus all others or high, moderate, low, none), PCM services (yes/no), a time indicator (1 18 months), and a time by PCM interaction. We used GEE cumulative logistic models for repeated measures with inpatient and emergency department utilization (number of claims ranged from 0 to 5) as the dependent variables. For outpatient and medical service utilization and all charge models, we used GEE normal regression methods for repeated measures. The dependent variable for medical service and outpatient utilization was number of claims (0 11 and 0 67, respectively). Log dollars was the dependent variable for all charge models. When compared with patients who were eligible but did not receive PCM services, those who received PCM services demonstrated no significant longitudinal differences at the 0.05 level in emergency department, inpatient, outpatient, or medical service utilization or charges (Table 6). The time by PCM P-value for medical service analysis was.086 (Table 6) when PCM claims were included and excluding PCM claims (not shown). To evaluate whether there was a dose response relationship between the intensity with which the service was adopted and effects on health care utilization, we examined the change in emergency department use, inpatient hospitalization, outpatient facility use, and medical/hcfa-1500 claims and charges using the same models as described earlier, testing for an intensity by time interaction. No significant intensity by time interaction was observed in any of the models at the 0.05 level (Table 6). Because PCM claims were included in the medical/hcfa-1500 claims file we continued our investigation of the effect of PCM on Vol. 44, No. 3 May/June Journal of the American Pharmacists Association 345

10 Iowa Pharmaceutical Case Management Table 5. Medications and Medicaid Charges on Index Date and at Follow-up Adjusted for Age and Gender a Patients Eligible But Did Who Received PCM Services Not Receive PCM Services (n = 524) (n = 1,687) Descriptor Index Date Follow-up P-value Index Date Follow-up P-value Active medications Mean (± SE) number of active drugs 7.5 ± ± ± ± Median number active drugs Range in number of active drugs By drug Mean (SE) amount billed per drug for active drugs ($) ± ± ± ± 2.07 <.001 Median amount billed per drug for active drugs ($) <.001 By patient Mean (SE) amount billed per patient for active drugs ± ± ± ± <.001 Median amount billed per patient for active drugs <.001 Abbreviations used: SE, standard error; PCM, pharmaceutical case management. a The index date was the day first eligible for PCM: follow-up was 9 months after the index date. No significant results for time by PCM interaction for any variables. This indicates that the change over time in these variables was the same for those who received and those who did not receive PCM services. Results were generated using generalized estimating equations for repeated measures with two time points for continuous ( normal regression) and binomial (logistic regression) variables. HCFA-1500 charges. We tested for effects in subgroups of patients. Patients were grouped according to characteristics known to influence receipt of PCM (age, gender, and number of medications). We totaled charges for 6 months before the index date and 12 months after the index date and adjusted the two totals for number of months. We then tested to see if, in any subgroup of patients, those who received PCM differed from those who did not receive PCM in their rate of change in HCFA-1500 charges. We again used GEE repeated-measures normal regression methods with log charges as the dependent variable at the two time points. We controlled for age and gender and tested for a significant time by PCM interaction to evaluate change over time. None of the 18 different subgroups had a statistically significant difference between those who received PCM and those who did not in patterns of monthly HCFA-1500 charges over time. Discussion The Iowa Medicaid PCM program was developed based on evidence demonstrating that collaboration between pharmacists and physicians improves medication safety Throughout the 2-year evaluation period, 1,440 billed services were provided by 114 pharmacies to 690 of 3,037 eligible patients. Physicians accepted 49.2% of pharmacist recommendations, and patients who received PCM services experienced significant improvements in medication use. The most common recommendation was to start a medication, and the most common reason was an untreated or undertreated condition. Discontinuation, changes in drug dose, and switches to alternate drugs also were common recommendations. Improvement in medication use was consistent across different medication problems. This is comparable with results found in a study by Hanlon et al. 3 who found that the percentage of medication problem ratings decreased during the study for 7 of 10 MAI categories in their intervention group. Also of interest is that the problem ratings increased for 5 of the 10 medication problems in the control group of that study. Our mean baseline MAI rating of 9.4 was somewhat lower than those in other studies of pharmacist interventions for high-risk patients for whom mean MAI scores have ranged from about 10 to 15. 3,30 32 The Iowa Medicaid PCM program intervention, which resulted in a mean change in MAI score of 1.1 points, appears to be less potent than institutional interdisciplinary team care. The typical change in MAI scores in prior studies has been approximately 4 or 5 points. Previous studies have used the MAI to evaluate interdisciplinary team interventions in institutional settings involving a small number of care providers. In contrast, we used the MAI to study pharmacist interventions in the context of busy dispensing pharmacies. The rate of physician acceptance of pharmacist recommendations (49%) was lower in this study than in hospital- or clinic-based collaborative practice settings, where acceptance rates have ranged from 83% to 95% Though smaller, our mean change in MAI score following PCM is probably clinically relevant. Schmader et al. 30 found that changes in total MAI scores of points were correlated with emergency department and hospital use and that a change of 1.7 points for cardiac medications was associated with improved blood pressure control. This is the first study to examine the effect of a community pharmacy physician collaborative intervention on the use of highrisk medications. Provision of PCM services was associated with a decrease in use of high-risk medications among older adults 346 Journal of the American Pharmacists Association May/June 2004 Vol. 44, No. 3

11 Iowa Pharmaceutical Case Management RESEARCH Table 6. Specifications and Results for Eight GEE Repeated-Measures Analysis Models of the Effects of PCM on Health Care Utilization Independent Variable Dependent Variable Interaction P-value Receipt (yes/no) of PCM a No. emergency department claims.340 Receipt (yes/no) of PCM a No. inpatient institutional claims.807 Receipt (yes/no) of PCM a No. outpatient facility claims.121 Receipt (yes/no) of PCM a No. medical/hcfa-1500 claims.086 Pharmacy intensity (high, moderate, low, zero) b No. emergency department claims.330 Pharmacy intensity (high, moderate, low, zero) b No. inpatient institutional claims.839 Pharmacy intensity (high, moderate, low, zero) b No. outpatient facility claims.112 Pharmacy intensity (high, moderate, low, zero) b No. medical/hcfa-1500 claims.616 Receipt (yes/no) of PCM a Charge of emergency department claims.513 Receipt (yes/no) of PCM a Charge of inpatient institutional claims.937 Receipt (yes/no) of PCM a Charge of outpatient facility claims.107 Receipt (yes/no) of PCM a Charge of medical/hcfa-1500 claims.064 Pharmacy intensity (high, moderate, low, zero) b Charge of emergency department claims.652 Pharmacy intensity (high, moderate, low, zero) b Charge of inpatient institutional claims.862 Pharmacy intensity (high, moderate, low, zero) b Charge of outpatient facility claims.212 Pharmacy intensity (high, moderate, low, zero) b Charge of medical/hcfa-1500 claims.166 Abbreviations used: GEE, generalized estimating equations; HCFA, Health Care Financing Administration; PCM, pharmaceutical case management. a Number of claims (or log charge) = age, gender, month (1 18), intervention (yes/no), month intervention. b Number of claims (or log charge) = age, gender, month (1 18), intensity (high, moderate, low, none), month intensity. (aged 65 and over) from 43.4% to 32.6%, representing a clinically substantial and statistically significant 24.9% improvement. This was in contrast to PCM-eligible patients who did not receive PCM, among whom the percentage of older adults taking high-risk medications did not change. Older adults who received PCM services had a higher prevalence of high-risk medication use than did those who did not receive the service. When groups differ at baseline in mean values of the variable of interest, a concern arises that the two groups may have had intrinsic differences in rate of change. This bias is termed selection maturation interaction. 37 As recommended by Cook and Campbell, 37 to rule out exogenous differing change rates as an explanation for between-group differences, we plotted monthly data from 3 months before through 9 months after initial PCM eligibility (Figure 1). The plotted lines did not differ in slope until almost 6 months after PCM eligibility. This postintervention change in slope following parallel baseline rates supports an intervention effect and argues against selection maturation interaction as an alternative explanation. Health care utilization patterns for patients who received PCM services were similar to those of patients who did not receive PCM services. Health care utilization was a secondary endpoint in this study, and we anticipated that this study would have insufficient power to detect small improvements in these measures, even though these can translate into substantial health and economic benefits. In addition, this was necessarily an evaluation of shortterm effects of PCM. Health status and health care utilization benefits likely require a longer time to be realized. In contrast with other studies, 8,38 we found no evidence that physician visits increased. We suspect that this can be attributed to the design of the program in which the pharmacist implements the action plan agreed upon after fax communication with the physician without requiring the patient to make a physician office visit. Despite these positive findings, the PCM program clearly experienced start-up challenges similar to those documented in other studies of community pharmacies. Of 3,037 eligible patients, only 690 patients received full PCM services (22.7%). The main challenges to establishing PCM services were related to patient access, pharmacist issues, physician awareness, and changing the existing systems of care. 22,39 The effort to start up this new service rested largely with the pharmacist. When a pharmacy received its list of eligible patients, a pharmacist contacted the patients, scheduled appointments, met with them, obtained additional information from their physicians if necessary, completed assessments, and forwarded written recommendations to physicians. Unlike the typical physician s office with staff to perform these types of activities, pharmacies lack such staff. Because of the time needed to complete all of these steps, several months may be required to finalize an action plan for a patient. Additional staff may be needed. In the face of uncertainty about the longevity of the PCM program and the effects of staffing changes on pharmacy finances, managers would be understandably reluctant to make such changes during the initial year of the program. Physician pharmacist communication obstacles were evident. Vol. 44, No. 3 May/June Journal of the American Pharmacists Association 347

12 Iowa Pharmaceutical Case Management Even though 49.2% of the pharmacists recommendations were accepted by physicians, lack of acceptance was not often a direct disapproval of the recommendation. Rather, physicians often did not respond (to either approve or disapprove) despite sometimes repeated communications. Physician systems may not have accommodated the program well, and physicians may not have been sufficiently aware of the program. Clearly, however, some of the pharmacists and physicians worked together very effectively. Our study of high-risk patients taking multiple medications documented a significant public health problem in a vulnerable population. These patients health status was far below the national average. They had an extremely high rate of medication problems. In adults aged 65 and older who received PCM services, 43.4% had been taking at least one medication considered to have a poor risk benefit balance and to be inappropriate for use among older adults. This rate is higher than published rates for community (22% 28%), general Medicaid (27%), and Medicaid nursing home (33% 40%) samples. 40 That these risk indicators identify people who suffer health consequences of inappropriate medication use is supported by the 26% prevalence of adverse drug reactions in the previous year. This rate is alarming and is nearly three times the rate observed in a different population of elderly Iowans not on Medicaid. 2 Limitations The program was implemented as a state plan amendment and the Iowa Department of Human Services determined that a randomized trial would not be allowed. The study design was therefore a nonrandomized comparison group design, with the comparison groups consisting of PCM-eligible patients who received the service and PCM-eligible patients who did not receive the service. Those who received PCM were taking more medications and had more medication safety problems at baseline, suggesting that pharmacists tended to select patients with the greatest risk. Supporting the conclusion that improvement in medication safety was attributable to PCM were the findings of dose response (pharmacy intensity) analyses. Future studies should strive to perform randomized controlled designs, but this is often difficult or impossible when a service such as the Iowa PCM program is implemented for an entire population. One individual conducted all chart-based MAI reviews. This should not be a serious limitation since individual items in the MAI have demonstrated excellent interrater reliability in previous work (kappa = 0.83 for physician internist agreement; kappa = 0.64 for two pharmacists) 26 and high interrater reliability has also been obtained for the MAI scores (intraclass correlation coefficient = 0.74). 27 Intrarater reliability of individual items was also high (kappa = 0.92). Conclusion PCM services were delivered to Iowa Medicaid patients at high risk of adverse medication experiences. Indeed, 26% of these patients reported experiencing adverse drug reactions in the year before the program, 2.6 times the rate in the general population of older Iowans, using the same survey instrument. In this report we have described the initial start-up experience with the Iowa Medicaid PCM program that was designed for these high-risk patients. A large number of patients received PCM services and medication safety improved for these patients. Because of the complexity of the program, the complexity of the patient population, and physicians general unfamiliarity with the PCM concept, the large number of patients who received care must be considered a success. Developing and sustaining PCM services in community pharmacy is clearly challenging. Implemented as a state plan amendment, the Iowa Medicaid PCM program continues to be available to Iowa s most vulnerable Medicaid patients. The beneficial effects observed among the patients who received these services in the start-up year support efforts to expand this program by enrolling more pharmacy providers, informing patients of their eligibility, raising physician awareness, and developing best practices models of community-based physician pharmacist collaborative PCM. References 1. Kohn LT, Corrigan JM, Donaldson MS, eds. To err is human: building a safer health system. Institute of Medicine Committee on Quality of Healthcare in America. Washington, D.C.: National Academy Press; Chrischilles EA, Segar ET, Wallace RB. Self-reported adverse drug reactions and related resource use. A study of community-dwelling persons 65 years of age and older. Ann Intern Med. 1992;117: Hanlon JT, Weinberger M, Samsa GP, et al. A randomized, controlled trial of a clinical pharmacist intervention to improve inappropriate prescribing in elderly outpatients with polypharmacy. Am J Med. 1996;100: Carter BL, Helling DK. Ambulatory care pharmacy services: has the agenda changed? Ann Pharmacother. 2000;34: McKenney JM, Slining JM, Henderson HR, et al. The effect of clinical pharmacy services on patients with essential hypertension. Circulation. 1973;48: McKenney JM, Brown ED, Necsary R, Reavis HL. Effect of pharmacist drug monitoring and patient education on hypertensive patients. Contemp Pharm Pract. 1978;1: Park JJ, Kelly P, Carter BL, Burgess PP. Comprehensive pharmaceutical care in the chain setting. J Am Pharm Assoc. 1996;36: Carter BL, Barnette DJ, Chrischilles E, et al. Evaluation of hypertensive patients after care provided by community pharmacists in a rural setting. Pharmacotherapy. 1997;17: Rupp MT, McCallian DJ, Sheth KK. Developing and marketing a community pharmacy-based asthma management program. J Am Pharm Assoc. 1997;37: Shibley MCH, Pugh CB. Implementation of pharmaceutical care services for patients with hyperlipidemias by independent community pharmacy practitioners. Ann Pharmacother. 1997;31: Bluml BM, McKenney JM, Cziraky MJ, Elswick RK Jr. Interim report from Project ImPACT: Hyperlipidemia. J Am Pharm Assoc. 1998;38: Journal of the American Pharmacists Association May/June 2004 Vol. 44, No. 3

13 Iowa Pharmaceutical Case Management RESEARCH 12. Currie JD, Chrischilles EA, Kuehl AK, Buser RA. Effect of a training program on community pharmacists detection of and intervention in drugrelated problems. J Am Pharm Assoc. 1997;37: Kuehl AK, Chrischilles EA, Sorofman BA. System for exchanging information among pharmacists in different practice environments. Consult Pharm. 1998;5: Ernst ME, Charlstrom CV, Currie JD, Sorofman B. Implementation of a community pharmacy-based influenza vaccination program. J Am Pharm Assoc. 1997;37: Ernst ME, Bergus GR, Sorofman BA. Patients acceptance of traditional and nontraditional immunization providers. J Am Pharm. Assoc. 2001;41: Hepler CD. Looking toward a market for pharmaceutical care. J Am Pharm Assoc. 1997;37: Miller LG, Scott DM. Documenting indicators of pharmaceutical care in rural community pharmacies. J Managed Care Pharm. 1996;2: Ellis SL, Carter BL, Malone DC, el al. Clinical and economic impact of ambulatory care clinical pharmacists in management of dyslipidemia in older adults: the IMPROVE study. Impact of managed pharmaceutical care on resource utilization and outcomes in Veterans Affairs Medical Centers. Pharmacotherapy. 2000;20: Gurwitz JH, Field TS, Harrold L, et al. Incidence and preventability of adverse drug events among older persons in the ambulatory setting. JAMA. 2003; 289: Chrischilles E, Sorofman B, Zieglowsky MB, et al. The Iowa on-line prospective drug utilization review (OPDUR) demonstration project: Study design, system design, and conceptual models. J Res Pharm Econ. 1997;8: Rovers JP, Currie JD, Hagel HP, et al. A practical guide to pharmaceutical care. Washington, D.C.: American Pharmaceutical Association; 1998: Carter BL, Chrischilles EA, Scholz D, et al. Extent of services provided by pharmacists in the Iowa Medicaid Pharmaceutical Case Management Program. J Am Pharm Assoc. 2003;43: Lund BC, Chrischilles EA, Carter BL, et al. Development of a computer algorithm for defining a cross-sectional drug regimen from electronic pharmacy claims. J Clin Epidemiol. 2003;56: Beers MH. Explicit criteria for determining potentially inappropriate medication use by the elderly. An update. Arch Intern Med. 1997;157: Ware JE, Kosinski M. SF-36 Physical & Mental Health Summary Scales: a manual for users of version 1. 2nd ed. Lincoln, R.I.: Quality Metric; Samsa GP, Hanlon JT, Schmader KE, et al. A summated score for the Medication Appropriateness Index: development and assessment of clinimetric properties including content validity. J Clin Epidemiol. 1994;47: Fitzgerald LS, Hanlon JT, Shelton PS, et al. Reliability of a modified medication appropriateness index in ambulatory older persons. Ann Pharmacother. 1997;31: Kassam R, Martin LG, Farris KB. Reliability of a modified medication appropriateness index in community pharmacies. Ann Pharmacother. 2003;37: Liang KY, Zeger SL. Longitudinal data analysis using generalized linear models. Biometrika. 1986;73: Schmader K, Hanlon JT, Landsman PM, et al. Inappropriate prescribing and health outcomes in the elderly in a pharmacist intervention trial. Ann Pharmacother. 1997;31: Jeffery S, Ruby CM, Hanlon JT, Twersky J. The impact of an interdisciplinary team on suboptimal prescribing in a long-term care facility. Consult Pharm. 1999;14: Hanlon JT, Schmader KE, Artz M, et al. The effect of geriatric evaluation and management units (GEMU) and clinics (GEMC) on drug-related problems (DRPs) [abstract]. Gerontologist. 2002;42(Spec No 1): Lobas NH, Lepinski PW, Abramowitz P. Effects of pharmaceutical care on medication cost and quality of patient care in an ambulatory-care clinic. Am J Hosp Pharm. 1992;49: Haxby DG, Weart CW, Goodman BW Jr. Family practice physicians perceptions of the usefulness of drug therapy recommendations from clinical pharmacists. Am J Hosp Pharm. 1988;45: Brown DJ, Helling DK, Jones ME. Evaluation of clinical pharmacist consultations in a family practice office. Am J Hosp Pharm. 1979;36: Bjornson DC, Hiner WO, Potyk RP, et al. Effect of pharmacists on health care outcomes in hospitalized patients. Am J Hosp Pharm. 1993;50: Cook TD, Campbell DT. Quasi-experimentation: design and analysis issues for field settings. Boston, Mass.: Houghton Mifflin; 1979: Malone DC, Carter BL, Billups SJ, et al. An economic analysis of a randomized, controlled, multicenter study of clinical pharmacist interventions for high-risk veterans: the IMPROVE study. Impact of managed pharmaceutical care resource utilization and outcomes in Veterans Affairs Medical Centers. Pharmacotherapy. 2000; 20: Chrischilles EA, Carter B, Voelker M, et al. Iowa Medicaid pharmaceutical case management program: report of the program evaluation. Iowa City: University of Iowa College of Public Health Technical Report #03-1; Liu GG, Christensen DB. The continuing challenge of inappropriate prescribing in the elderly: an update of the evidence. J Am Pharm Assoc. 2002;42: Hansten PD, Horn JR. Hansten and Horn s drug interactions, analysis and management. Vancouver, Wash.: Applied Therapeutics; PHARMACY THROUGH THE AGES Lives Interrupted In 1942 almost 120,000 Japanese Americans were removed from the West Coast. Sent first to hastily established assembly centers in racetracks and fairgrounds, they were eventually moved to 10 relocation camps in hostile and inhospitable locations in the interior. Pharmacy students and pharmacists were among those dispossessed. Students were removed from the University of Washington, Oregon State, University of California San Francisco, and the University of Southern California. Pharmacy owners were forced to dispose of their businesses within a few short weeks. A November 1942 survey noted that almost 100 pharmacists were working in the camp hospitals and satellite clinics. A photo of the hospital pharmacy at the Jerome Relocation Camp, Denson, Arkansas, shows pharmacist Yaeko Mayedo labeling a prescription. Some of the relocation camps remained in operation until late SOURCE: NATIONAL ARCHIVES PHOTOGRAPH NWDNS-210-G-E792, TOM PARKER, PHOTOGRAPHER. Dennis B. Worthen, PhD, is Lloyd Scholar, Lloyd Library and Museum, Cincinnati, Ohio, and JAPhA Contributing Editor, Heroes of Pharmacy. Vol. 44, No. 3 May/June Journal of the American Pharmacists Association 349